Overview

The Vectara MCP Server provides agentic AI applications with access to fast and reliable Retrieval-Augmented Generation (RAG) capabilities, powered by Vectara's Trusted RAG platform through the Model Context Protocol (MCP). It's compatible with any MCP client, including Claude Desktop.

Official Server:

Developed and maintained by Vectara

Key Features

⚡

Fast \u0026 Reliable RAG

Access to fast and reliable Retrieval-Augmented Generation with reduced hallucinations.

🔍

Semantic Search

Perform advanced semantic searches without generating responses.

🛡️

Hallucination Correction

Identify and correct hallucinations in generated text using Vectara's VHC API.

🔒

Secure \u0026 Scalable

Built-in authentication, HTTPS readiness, rate limiting, and CORS protection.

Available Tools

Quick Reference

Tool	Purpose	Category
`setup_vectara_api_key`	Configure and validate Vectara API key	API Key Management
`clear_vectara_api_key`	Clear stored API key	API Key Management
`ask_vectara`	Run RAG query with generated response	Query
`search_vectara`	Run semantic search query	Query
`correct_hallucinations`	Identify and correct hallucinations	Analysis

Detailed Usage

setup_vectara_api_key▶

Configure and validate your Vectara API key for the session (one-time setup).

use_mcp_tool({
  server_name: "vectara",
  tool_name: "setup_vectara_api_key",
  arguments: {
    api_key: "your_vectara_api_key"
  }
});

Returns success confirmation with masked API key or validation error.

clear_vectara_api_key▶

Clear the stored API key from server memory.

use_mcp_tool({
  server_name: "vectara",
  tool_name: "clear_vectara_api_key",
  arguments: {}
});

Returns confirmation message.

ask_vectara▶

Run a RAG query using Vectara, returning search results with a generated response.

use_mcp_tool({
  server_name: "vectara",
  tool_name: "ask_vectara",
  arguments: {
    query: "What is Vectara?",
    corpus_keys: ["my-corpus"],
    n_sentences_before: 2,
    n_sentences_after: 2,
    lexical_interpolation: 0.005,
    max_used_search_results: 10,
    generation_preset_name: "vectara-summary-table-md-query-ext-jan-2025-gpt-4o",
    response_language: "eng"
  }
});

Returns the response from Vectara, including the generated answer and the search results.

search_vectara▶

Run a semantic search query using Vectara, without generation.

use_mcp_tool({
  server_name: "vectara",
  tool_name: "search_vectara",
  arguments: {
    query: "What is Vectara?",
    corpus_keys: ["my-corpus"],
    n_sentences_before: 2,
    n_sentences_after: 2,
    lexical_interpolation: 0.005
  }
});

Returns the response from Vectara, including the matching search results.

correct_hallucinations▶

Identify and correct hallucinations in generated text using Vectara's VHC (Vectara Hallucination Correction) API.

use_mcp_tool({
  server_name: "vectara",
  tool_name: "correct_hallucinations",
  arguments: {
    generated_text: "The generated text to analyze for hallucinations.",
    documents: ["document1", "document2"]
  }
});

Returns the analysis of the generated text with hallucination corrections.

Installation

{
  "mcpServers": {
    "vectara": {
      "command": "python",
      "args": [
        "-m",
        "vectara_mcp"
      ],
      "env": {
        "VECTARA_API_KEY": "your_vectara_api_key"
      }
    }
  }
}

Custom Configuration:

Replace your_vectara_api_key with your actual Vectara API key. Additional environment variables like VECTARA_AUTHORIZED_TOKENS, VECTARA_ALLOWED_ORIGINS, VECTARA_TRANSPORT, and VECTARA_AUTH_REQUIRED can be configured.

Common Use Cases

1. Enhancing AI applications with reliable data retrieval

use_mcp_tool({
  server_name: "vectara",
  tool_name: "ask_vectara",
  arguments: {
    query: "What are the benefits of using Vectara for RAG?",
    corpus_keys: ["product-documentation"]
  }
});

2. Running complex queries to extract information from various data sources

use_mcp_tool({
  server_name: "vectara",
  tool_name: "search_vectara",
  arguments: {
    query: "Show me all documents related to 'AI ethics' from the last quarter.",
    corpus_keys: ["research-papers", "news-articles"]
  }
});

3. Integrating with other AI systems for improved performance

// Example: A chatbot uses Vectara to answer user questions
const userQuery = "How does Vectara handle data privacy?";
const vectaraResponse = use_mcp_tool({
  server_name: "vectara",
  tool_name: "ask_vectara",
  arguments: {
    query: userQuery,
    corpus_keys: ["privacy-policy", "legal-documents"]
  }
});
// Further processing of vectaraResponse by the chatbot

4. Hallucination correction

const generatedText = "Vectara is a company that sells shoes.";
const sourceDocuments = ["Vectara is a company that provides a RAG platform.", "Vectara focuses on AI and search."];
use_mcp_tool({
  server_name: "vectara",
  tool_name: "correct_hallucinations",
  arguments: {
    generated_text: generatedText,
    documents: sourceDocuments
  }
});

Connection String Format

The Vectara MCP server does not use a traditional connection string. Instead, it relies on environment variables for configuration and API keys for authentication. The server runs as an HTTP or SSE service.

API Key: Set VECTARA_API_KEY environment variable.
Authentication: Bearer tokens are used for HTTP/SSE transport.
Host/Port: Configured via command-line arguments (--host, --port) or defaults to http://127.0.0.1:8000.

Vectara MCP Server